iteration Vs LUT table entry vs accuracy in Cordic

[praveen]

Hello,
I wanted the relationship between the number of iteration vs LUT table
entry.
Because i wanted to estimate the arc tan to a accuracy of 1
microradians.
In my simulation i found that even if the iteration is around 25
iteration i could not achieve the 1 microradian accuracy.

My LUT i have represented using 32 bit and the two number whos atan is
to be obtained is also 32 bit(all fixed point).

Please suggest a solution
with regards
praveen

 

[glen herrmannsfeldt]

praveen wrote:

I wanted the relationship between the number of iteration vs LUT table
entry.
Because i wanted to estimate the arc tan to a accuracy of 1
microradians.
In my simulation i found that even if the iteration is around 25
iteration i could not achieve the 1 microradian accuracy.

Somewhere I have a databook from many years ago with a ROM look-up
table for sine. This was when ROMs were much smaller than today,
and pretty much the way it worked was to have one that was the
high bits of the sine, and another that did linear interpolation.

What domain and range do you expect for your arctan?
For smaller values it shouldn't be too hard, but for larger
values it gets much harder. What form is the input in, and what
form do you expect the output to be?

-- glen

 

[Jerry Avins]

praveen wrote:

Hello,
I wanted the relationship between the number of iteration vs LUT table
entry.
Because i wanted to estimate the arc tan to a accuracy of 1
microradians.
In my simulation i found that even if the iteration is around 25
iteration i could not achieve the 1 microradian accuracy.

My LUT i have represented using 32 bit and the two number whos atan is
to be obtained is also 32 bit(all fixed point).

Please suggest a solution
with regards
praveen

Go to my web site, http://users.erols.com/, look at my article
"Quadratic Interpolation in Forth", and even it the article itself
gives you no ideas, there's a link at the top that may.

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

 

[Ken Prager]

In article <40280492$0$3199$61fed72c@news.rcn.com>,
Jerry Avins <jya@ieee.org> wrote:

praveen wrote:

Please suggest a solution
with regards
praveen

Go to my web site, http://users.erols.com/, look at my article
"Quadratic Interpolation in Forth", and even it the article itself
gives you no ideas, there's a link at the top that may.

That link seems to be brokken--it redirects to <http://www.rcn.com/>.

KP

--
Remove _me_ for e-mail address

 

[Jerry Avins]

Jerry Avins wrote:

praveen wrote:

Hello,
I wanted the relationship between the number of iteration vs LUT table
entry.
Because i wanted to estimate the arc tan to a accuracy of 1
microradians.
In my simulation i found that even if the iteration is around 25
iteration i could not achieve the 1 microradian accuracy.

My LUT i have represented using 32 bit and the two number whos atan is
to be obtained is also 32 bit(all fixed point).

Please suggest a solution
with regards
praveen


Go to my web site, http://users.erols.com/, look at my article
"Quadratic Interpolation in Forth", and even it the article itself
gives you no ideas, there's a link at the top that may.

Jerry

Copy/paste mysteriously failed. (Operator error, but how?) The correct
URL is http://users.erols.com/jyavins/ or http://users.rcn.com/jyavins/
or even http://users.starpower.net/jyavins/

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ